Leveraging back-off grammars for authoring context-free grammars

• US 8,108,205 B2
• Filed: 12/01/2006
• Issued: 01/31/2012
• Est. Priority Date: 12/01/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Paek, Timothy, Chickering, Max, Badger, Eric
• Primary Examiner(s): Hudspeth, David R
• Assistant Examiner(s): SPOONER, LAMONT M
• Application Number:
  US11/607,284
• Publication Number:
  US 20080133220A1
• Time in Patent Office: 1,887 Days
• Field of Search: 704/9, 704243-244, 704/251, 704/257
• US Class Current: 704/9
• CPC Class Codes: G06F 40/211 Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars

Abstract

A system and method of refining context-free grammars (CFGs). The method includes deriving back-off grammar (BOG) rules from an initially developed CFG and utilizing the initial CFG and the derived BOG rules to recognize user utterances. Based on a response of the initial CFG and the derived BOG rules to the user utterances, at least a portion of the derived BOG rules are utilized to modify the initial CFG and thereby produce a refined CFG. The above method can carried out iterativey, with each new iteration utilizing a refined CFG from preceding iterations.

Claims

1. A speech recognition method comprising:

recognizing, by a processor, a user utterance using a context-free grammar (CFG) as a language model;

identifying, by the processor, out-of-grammar (OOG) utterances that are not recognizable by the CFG;

processing, by the processor, the OOG utterances through a back-off grammar (BOG) that is derived from the CFG and that includes a plurality of BOG rules;

collecting statistics related to a degree of success of the plurality of BOG rules in processing the OOG utterances and statistics related to a frequency of each of the plurality of BOG rules being considered a top recognition result;

utilizing a dictation grammar to selectively transcribe a portion of the OOG utterances based at least in part on the collected statistics; and

incorporating the transcribed portion of the OOG utterances into the CFG.

2. The method of claim 1 wherein the transcribed portion of the OOG utterances is automatically incorporated into the CFG.

3. The method of claim 2 and further comprising repeating the steps of claim 2 iteratively until an acceptable speech recognition efficiency level is reached.

4. The method of claim 1 and further comprising saving the portion of the OOG utterances as file copies.

5. The method of claim 4 wherein the portion of the OOG utterances are saved as .wav files.

6. The method of claim 1 wherein the transcribed portion of the OOG utterances is manually incorporated into the CFG.